Generating minimal living systems from non-living materials and increasing their evolutionary abilities

Abstract

We review lessons learned about evolutionary transitions from a bottom-up construction of minimal life. We use a particular systemic protocell design process as a starting point for exploring two fundamental questions: (i) how may minimal living systems emerge from non-living materials? and (ii) how may minimal living systems support increasingly more evolutionary richness? Under (i), we present what has been accomplished so far and discuss the remaining open challenges and their possible solutions. Under (ii), we present a design principle we have used successfully both for our computational and experimental protocellular investigations, and we conjecture how this design principle can be extended for enhancing the evolutionary potential for a wide range of systems.

This article is part of the themed issue ‘The major synthetic evolutionary transitions’.

1. Introduction

Novel functionalities in physico-chemical systems can be generated naturally in three ways: (i) by the assembly of structures (equilibrium processes), (ii) by self-organization (non-equilibrium processes), and (iii) by a combination of the two, through the evolution of structures [1,2].

Our approach to create minimal living systems, which we define as protocells [3–15], uses coupled self-assembling and self-organizing processes. We investigate how a controlled environment together with coupled self-assembly and externally driven self-organization may play together to generate minimal, self-replicating, physico-chemical systems.

Minimal life is notoriously difficult to define as the systems we study straddle a grey zone from processes with some life-like behaviours to what most scientists agree to be fully living systems. However, for physico-chemical systems there is consensus on an operational definition of minimum life based on three interconnected functionalities, a metabolism, an informational system and a container, which in a controlled environment are able to harvest external free energy to convert resources into building blocks so that the system can grow and divide, in part, controlled by inheritable information. Further, as the information can change from one generation to the next, different information may generate different growths and division patterns so that selection and thus evolution are possible [16]. For details see figures 1 and 2.

Figure 1.

(inner triangle) A system of three interconnected components, a container that connects a metabolism and an informational system, which in a given environment, can transform resources into building blocks, grow, divide and undergo evolution (outer circle). Detailed discussions follow in §§2 and 3.

Figure 2.

Obtaining organizational and functional closure is a key issue to construct a protocell bottom up. Resources and free energy are provided from the right side and are part of the environment. The free energy is used to process or ‘digest’ the resources and turn them into building blocks (dG > 0), which defines the primitive metabolism. The metabolism can be light-driven or fuelled by chemical redox energy. One set of building blocks self-assemble into a container (dG < 0), which, for example, can be a vesicle or a droplet. Another set of the building blocks is used to construct new informational molecules. One function of the container is to keep the metabolism and the informational system in close proximity, which is done through self-assembly as the metabolic and informational components, for example, have hydrophobic anchors so they attach to the container (surface of a vesicle) or through container encapsulation (inside a vesicle). The informational system facilitates the chemical production of the building blocks, which means the system increases the metabolic rate kinetics through catalysis. Further, the information complex can make copies of itself from the building blocks produced in the metabolic process, such that catalytic capabilities can be inherited from one generation to the next.

2. A minimal protocell

Figure 2 shows the functional and organizational design of a minimal living system. Our systemic protocell assembly approach requires a simple metabolism (i.e. few very simple catalytic processes) that is controlled by information, e.g. serving as an electron relay, where both are kept together by a container.

Over the years our team has successfully implemented a protocell around a ruthenium tris(bipyridine) [Ru(bpy)₃] complex that uses light to catalyse redox reactions on precursors of both the amphiphiles and the information in bulk (figure 3). The informational system serves as part of an electron relay that modulates the metabolic reaction rate, which in turn depends on the redox potential (the nucleobase composition) of the information molecule.

Figure 3.

Metabolism at the centre of the protocell design. (a) Production of amphiphiles. (b) Ligation of the DNA between a 3′-phosphoimidazole phosphate and a 5′-picolyl protected 5′-amino oligomers. Step I: in both cases, the photocleavage of a picolyl group by the ruthenium complex [box: 8-oxoG-Ru(bpy)₃ complex] delivers the protocell building blocks. In (b), two subsequent steps take place: steps II: decarboxylation and III: formation of the phophorimidate bond. (c) Scheme of the photochemistry. Upon irradiation, a Ru MLCT is formed that alone cannot perform the photoclevage. Upon transfer of an electron from the information molecule (8-oxoG), a functional reductant is formed that converts the amphiphile precursor (pL) into an amphiphile (L⁻). A sacrificial H-atom donor is used to regenerate the 8-oxoG and scavenge the radical. (d) Impact of the information-photocatalyst on the conversion rates. Red and black, 8-oxoG-Ru(bpy)₃ with a hydrocarbon moiety and without it (aqueous complex), respectively. In blue, the catalysis by a guanine Ru complex. All samples were the same except for the catalyst. For more details and references, see text.

In particular, we have established that the amphiphile production can be controlled by chemical information. The reduction potential of a nucleobase, 8-oxo-guanine [oxoG], can be exploited by the photocatalyst to produce amphiphiles, but not that of guanine (the next most easily oxidized nucleobase) or by extension, those of A, C, U and T [17]. Furthermore, fatty acid vesicles will influence the production rates as a detailed investigation of the information-photocatalyst configuration showed, especially when both oxoG and the Ru(bpy)₃ are attached to the container via hydrophobic anchors [18]. Further, we have established a photochemical fragmentation scheme to ligate DNA oligomers. First, the deprotection of an oligomer is performed using a Ru(bpy)₃ photosensitizer. This oligomer can only then, and in the presence of a template, be ligated with another oligomer [19]. Further, we have demonstrated that DNA conjugated bola-amphiphile can be anchored to a fatty acid vesicle [20], demonstrating how information and container can be linked as well as how metabolically driven (Ru(bpy)₃ photosensitizer) self-replicating vesicles indeed divide through budding and thus inherit anchored metabolic molecules [21,22] (figure 3).

Our simulation studies have focused on both detailed processes [16] and full protocell life-cycle simulations, where one study demonstrates how a catalytic coupling between the container and information growth results in an orchestrated overall exponential protocellular growth rate [23,24]; and another demonstrates the intricate interdependencies and challenges of integrating the timescales between the different components and processes [25].

Many other groups have addressed these same, or very similar, issues regarding container growth and division [26,27], formation of autocatalytic metabolic sets [28–30], information replication [13,31,32], thermodynamics of replication [33,34], protocellular integration [11,13], as well as coupling and synchronization of growth among the different protocellular processes [35].

To summarize: we have experimentally demonstrated several advantages of our systemic approach integrating the three mutually supporting components in a controlled environment: (i) self-assembly of a decanoic acid container; (ii) anchoring to the container a metabolic ruthenium complex as well as (iii) a conjugated nucleic acid information complex; (iv) container feeding and growth; (v) metabolically driven container replication; (vi) metabolically driven information ligation (part of replication); (vii) one-pot metabolic production of both amphiphilic molecules and ligated oligomers, new information molecules. These are all key milestones towards the construction of a minimal living system. One key milestone is not yet reached, however, before full protocell integration can occur: implementation of an effective DNA self-replication process based on template-directed ligation of two smaller oligomers (figure 4).

Figure 4.

Overview of the implementation of our experimental systems together with envisioned systems integration. Enumerated subsystems 1–5 are implemented and published. The open challenges for implementing subsystem 6 are explored in the following §3. A more thorough discussion of minimal life together with different kinds of evolution is presented in §4.

From our discussions above, we have emphasized that information replication has not yet been implemented. Thus, we need to develop a template-directed nucleotide ligation with a subsequent robust strand-replication process.

3. Information replication, catalysis, inheritance and evolution in a protocell

In the following, we present a theoretical review of our key open challenge for implementing an integrated functional protocell: how to optimize and realize an independent and robust replication of the protocellular information molecules? We address this problem independently of the issues associated with nucleotide strand attachment to the protocellular container surface. This simplifies the problem and we believe the replication issues can be treated independently from the strand attachment issues, which we have resolved experimentally [20]. Our analysis seeks to determine how to optimize the overall information replication rate dependent on three key experimental parameters: the hybridization energies, the information strand length and the reaction temperature. Subsequently, we discuss what the implementation of an information replication process means for the evolutionary possibilities of the protocells.

(a). Non-enzymatic, template-directed information replication

We now investigate the dependence of the overall replication rate constant on hybridization energies, temperature and strand length by employing a model for the minimal ligation-based replication process of a single-stranded template in which the ligation of oligomers is involved in the formation of the complementary replica. Within the template-directed replicator system, two complementary oligomers hybridize to a single-stranded template. An irreversible ligation reaction (i.e. formation of covalent bonds in a condensation reaction) transforms the oligomers into the complementary copy of the template. The newly formed double strand can dehybridize, thus allowing for iteration of the process. Throughout the replication mechanism, we neglect both the production of waste and the hydrolysis of ligation [36]. The overall reaction is summarized below:

where X and O denote the template strand and monomers/oligomers, respectively (figure 5).

Figure 5.

Template-directed ligation and product inhibition replication where X and O denote the template strand and oligomers, respectively.

The equilibrium constants corresponding to the reaction mechanism described above are given by

3.1

where i labels the type of reaction (i.e. hybridization O, ligation L and dehybridization T), are rate constants for the corresponding forward and backward reactions and ΔG_i denotes the free energy changes.

The differential equations expressing the rates of formation of each of the X-containing species are then

3.2

3.3

3.4

3.5

and the total molar concentration of the template is

3.6

Mass action kinetics under quasi-steady-state approximations [37]; Stadler et al. [38–40] translates the reaction mechanism into the following growth law

3.7

where α > 0 is the replication rate constant in solution and β > 0 is a measure for the product inhibition. The rate function Ψ is monotonically decreasing and can be written in the following form [41]

3.8

It satisfies Ψ(u) → 1 for small u and asymptotically behaves like for large u, where u indicates the total molar concentration of the template. The parameters α and β can be explicitly determined in terms of the elementary rate constants.

Template-directed replication generally suffers from product inhibition where most of the templates are in double-strand configuration. Throughout the reaction, this is equivalent to Under this assumption, equation (3.7) simplifies to [36]

3.9

where the oligomers concentration [O] is kept constant. Therefore, the template-directed replicator system undergoes the well-known parabolic growth with an overall replication rate constant k expressed in terms of equilibrium and elementary rate constants as

3.10

If the ligation step is rate limiting (i.e. ligation step is slow compared with the other reactions), the condition for the rate constants is satisfied. Under this assumption, the overall growth rate given by equation (3.10) reduces to

3.11

This is identical with the result obtained by Fellermann & Rasmussen [36] employing thermodynamic arguments and a polymer model for oligonucleotides that allows simulation of their diffusion and hybridization behaviour. The important observation of equation (3.11) is that the overall growth rate is independent of association and dissociation rates but depends on the equilibrium constants K_O and K_T.

If we, by contrast, assume that the ligation step is not rate limiting (i.e. ligation step is fast in comparison with the other reactions) then and from equation (3.10), the overall replication constant becomes

3.12

In contrast, assuming that our template-directed replication mechanism does not experience product inhibition, then the total template concentration satisfies and the dynamics of equation (3.7) is, therefore, described by a replicator equation of the form

3.13

In this case, the template-directed replicator system attains exponential growth, which promotes Darwinian evolution and the selection of the fittest (highest growth rate). Nevertheless, this situation is beyond the scope of our analysis.

The temperature and strand-length dependence of the overall replication rates given by equations (3.11) and (3.12) is emphasized in figure 6 under the conditions investigated above. Within the plots, the ‘temperature’ labels a quantity T′/T, in which T′ denotes a temperature scale and T indicates the actual temperature of the system.

Figure 6.

Effective overall replication rate constant k as a function of strand length and temperature. In (a), the template-directed replication mechanism is subjected to product inhibition and slow (i.e. rate-limiting) ligation reaction (see Eq. (3.11)) [36]. In (b), the replication mechanism is also subjected to product inhibition but exposed to a fast (i.e. not rate-limiting) ligation reaction (see Eq. (3.12)] [42].

In the case of a replication mechanism, which is subjected to product inhibition and slow ligation reaction (i.e. Figure 6a), below a critical strand length, the rate constant k decreases with decreasing temperature T (increasing T′/T). For strands longer than the critical length, the replication rate grows with decreasing temperature. For the rate-limiting ligation process, the longer the oligomers stay on the template the larger the probability that a ligation reaction will occur. Therefore, it is obvious that larger oligomer strand length or lower temperature increase the overall template production rate.

In the case of product inhibition, if the ligation process is not rate-limiting (i.e. figure 6b), then it will be the hybridization/dehybridization rates of the oligomers and ligated oligomers that determine the overall template production rate. Since the template binding energy is approximately twice that of the oligomers (independently of strand length), the hybridization/dehybridization rates will not be that different for different strand lengths, so that not much happens along that axis; whereas for high temperatures the oligomers will not hybridize and at low temperatures the ligated oligomers will not leave the template, hence there must be a critical temperature as shown.

Summary: The protocell inherits its information through the replication processes discussed above. The details of the sequence in terms of 8-oxo-guanine content and location, which define the compositional strand information, are assumed to catalyse the metabolic processes as 8-oxo-guanine and Ru(bpy)₃ were documented to do in §2.

The qualitative dynamics of the expected overall information replication rate is intricately dependent on the relative rates (constants) of the sub processes involved, which require further assumptions about the nucleobase sequence details involved as well as the environment [42].

In this short review, we have only included reactions that involve product inhibition because we believe it is the most likely scenario as we only use very short nucleobase systems. The experimental DNA templates we have used so far have all been shorter than 20 nucleobases. Also, the short nucleotide systems used by von Kiedrowski and co-workers [31,43] are dominated by product inhibition. For significantly longer RNA templates (including designed stem-loops), Joyce and co-workers [44] have demonstrated how a nucleotide system can escape product inhibition through a cleverly designed stem-loop system. Their system shows self-sustained evolution through exponentially growing strand multiplication.

(b). Limited protocellular evolvability in a constrained environment

In the presented protocellular system, evolution is defined in the following way: compositional information, which is defined as the content and location of 8-oxo-guanine in the information strand, determines the metabolic reaction rates through an electron relay. The electron relay consists of 8-oxo-guanine to Ru(bpy)₃ to the resource molecules (picolinium esther and picolil-protected oligomers), which are, respectively, converted into decanoic acid (container building blocks) and active oligomers that can be ligated into full nucleic acid strands (information systems building blocks). These processes require a variety of environmental conditions including sacrificial proton and electron donors, recall §2 [17–19]. Thus, a modification of the compositional information generally results in modified metabolic reaction rates.

Variation of the metabolic reaction rates across a protocellular population would mean different growth and division rates and thus different overall replication rates for different protocells. Thus, different protocellular replication rates open up for protocellular selection, and if the process is iterated, it results in Darwinian evolution, recall the theoretical documentation for exponential growth in §2 [23,24]. Such a process is expected to result in protocellular populations with increasing replication rates.

Thus, for the simple protocellular model in a fixed environment, we expect Darwinian evolution to be a metabolic reaction rate optimization process, where presumably the overall replication rate (the phenotype) is enhanced. This means a change at the molecular (genotypic) level, through selection and amplification of the compositional information of the nucleotide strands, as these strands are being inherited. Expressed in a different manner: in a constant environment, we do not anticipate significant evolutionary ‘discoveries’, e.g. in terms of novel reaction pathways, novel component architectures, novel energy sources.

Bedau et al. [45] proposed a statistical characterization of evolutionary processes that aims to quantify the innovative potential of an evolutionary process by measuring the rate at which innovative changes are produced during the evolutionary process. In this classification scheme our protocellular systems falls into Class 2, which includes optimization process. Class 1 is a neutral evolutionary process and includes diffusion processes, while Class 3 is defined as evolutionary processes with an apparent open-ended ability to innovate. Such processes include examples from biological evolution and technological evolution. For a detailed discussion of the current thinking on open-ended evolution, please see [46].

As part of the scientific community identifies open-ended evolution as a central property of living systems, the presented protocell model (or for that matter, any published protocellular model we are aware of) does not qualify as an example of a living physico-chemical process. However, if protocellular evolution of Class 2 suffices, several of the published protocellular models, if successfully integrated and implemented, would qualify as examples of minimal life forms.

4. Novelty from the environment and from encoded information

(a). Systemic design by carefully coupled self-assembly and self-organization

We have in the previous sections argued that a simple protocell could emerge from interacting constituent components through coupled self-assembling and self-organizing processes. However, we have also argued that replication, variation and selection of such a protocellular population system seems to have a limited evolutionary potential. Evolution would probably be limited to a rate optimization of the metabolic reaction, which in turn is caused by changes in the compositional information (position and amount of oxo-G).

A central question is therefore: how may constituent components of a system and its environment need to be designed to facilitate a more innovative evolutionary process?

We know from experimental experiences as well as theoretical investigations that constituent components and environment are too simple, only trivial emergent structures will be generated, and as appropriate diversity/complexity of the constituent components is increased, emergence of hierarchies or multilevel structure may occur [1,2,46].

We now discuss a simple theoretical formulation of our protocellular system to illustrate the connection between constituent environment/component properties and how their interactions may or may not generate interesting observable functionalities through their dynamics. A graphical representation is discussed in figure 7.

Figure 7.

Connection between the details included in the simulations and the ability of the simulations to generate targeted observables. Left side of table summarizes the included physical model (each row). Right side of table indicates the higher order observable phenomena/functionalities generated by the simulation. Top of table depicts the qualitative information details needed in a molecular model representation (the data structure) of the simulation (columns). Simulations with more detailed and thus complex molecular components are able to generate increasingly more complex dynamics and functionalities. As an example, data structure D₃ has included enough molecular interaction details to allow the simulation to generate molecular self-assembly and, for example, micellar and vesicle formation. The last row is left open as we conjecture that to obtain a higher evolutionary potential we need to add more components and resources to the system. More detailed discussion in §4a,b.

As we construct a simulation of the interacting molecular components, we need to decide how many molecular details and interactions to include [47,48]. A simplistic representation of a molecular component may only include an excluded volume interaction potential, which enables simple fluid dynamics simulations, e.g. as in hard billiard or lattice gas simulations [49] as well as in liquid noble gases interacting via the Lennard-Jones potential [50]. If we implement two different molecules by two different kinds of interaction potentials it becomes possible for us to simulate and observe fluid phase separations (e.g. oil and water). Implementation of molecular polymers in simulation requires an implementation of molecule-to-molecule bonds, which, for example, allows us to observe polymer elasticity [51,52]. If we implement two different kinds of monomers in the polymer, for example, one being hydrophobic and the other hydrophilic may allow us to observe the formation of micelles and membranes in simulation [53–55].

Further, if we implement chemical reactions between two molecules [56–58], e.g. catalysed by a third molecule, we can simulate simple metabolic reactions and, for example, observe micellar division [25,59,60]. Template-directed ligation replication may be generated in simulation if appropriate base recognition potentials are implemented as the informational molecules are anchored to a droplet surface [25,61,62].

It should be noted that all component interactions and processes involved in our proposed protocellular system have been implemented in three-dimensional physics-based simulations. However, a fully integrated simulation of the protocellular life cycle processes has not yet been obtained due to the challenges involved in combining the multiple time and length scales, which range from less than a microsecond (photo fragmentation and self-assembly) to hours (metabolically driven container replication).

In modern biology, evolutionary processes are usually observed as caused by variations in the DNA. Different DNA results in different RNA and proteins, because DNA encodes functionalities through a ribosome translation. In a system with encoded information, a change in the information strand may be considered neutral unless/until it is translated where the change may cause functional changes in the translated RNA/protein and thus the overall functionalities. Note that the discussed protocellular systems in the previous sections do not have such an information translation system. In these protocellular systems, a change in the informational system (e.g. DNA or RNA) may result in functional changes directly through the changes in catalytic properties of informational system (e.g. oxoG content and location or folding). Thus, functional changes may occur due to changes in the replicating nucleotide strand without the existence of an information-encoding system (ribosomes) as in modern biology.

(b). Larger evolutionary potential by enriching the environment

Our simulation and our experimental experiences indicate the following: more variability (more and different components in a richer environment) together with more properties/complexity in the molecular component interactions (more delicate interactions that, for example, differ depending on the context) can open up for richer observable functionalities in the dynamics. Thus, it is tempting to assume this could also be extended after self-replication, and simple Darwinian evolution is achieved for a protocell (figure 8).

Figure 8.

For a fixed environment an increased environmental richness through more available building blocks and energy sources will eventually be necessary to open up for more evolutionary possibilities. We conjecture that increasing environmental richness is a necessary condition for eventually obtaining open-ended evolution.

We here refer to two different kinds of novelty. The first kind of observable novelty emerges as a result of self-assembling and self-organizing processes in a fixed environment. The second kind of observable novelty stems from an expansion of different available resource components and energy sources in the original environment.

If we assume a fixed environment and if we assume that after a given period of time all possible system configurations have been visited, then it is necessary to extend the number of combinatorial states (the environment) to expand the configuration space. This is a tautology and it is, therefore, trivially true.

Many authors propose similar ideas and use an extending of the available resource set to obtain richer dynamics/evolution (e.g. [26,28,63]).

In practice, how could we implement more variation in the components and the controlled protocellular environment? More variation could include more and different resource oligomers (short nucleotide libraries), changes in the fatty acid composition and addition of different photosynthesizes molecules. This could be done stepwise by adding/integrating novel container and metabolic component materials into the protocellular system. Impact or performance could then be measured at the resulting metabolic rate, container-division properties and life-cycle (generation) time. However, given our experimental experiences, this would be a challenging and time-consuming enterprise as each new component in the mix in principle (easily) could cause undesired (destructive) side effects.

The expansion of the environment has to be carefully designed. Although the following discussion is hypothetical, it is concrete and specific in the manner by which we propose to expand the evolutionary potential of our protocellular populations: the informational molecule of the protocell could be diversified from its original form through sequence diversity. Each sequence variant may, therefore, have an impact on the performance of the protocell, primarily through how the reactive Ru(bpy) and 8-oxo-guanine are positioned relative to each other. One could exploit the combinatorial potential of DNA polymers by introducing a synthesized DNA oligo library into the system with up to 10¹³ variants. The DNA library could then be introduced into the protocell system and fast-replicating variants isolated and amplified. In such a SELEX procedure, novel DNAs could be isolated and characterized [64]. In addition, unanticipated intermolecular interaction among DNAs and between the DNAs and protocells container/photosensitizer may also add to the novelty of the system.

Single-chain amphiphiles with a medium chain length (C ≥ 10), e.g. fatty acids, can self-assemble into bilayers that form the boundaries of vesicles. We have expanded our studies to mixtures of these amphiphiles either with different hydrocarbon chains (e.g. monocarboxylic acids C2–C10, [65]) or with different head groups (ammonium, sulfate, sulfonate, phosphate, etc.). These mixed amphiphile systems often tend to form more stable structures, having different biophysical properties (e.g. permeability), than those observed when only one single amphiphile type is used [21,22]. We have even demonstrated that a single-chain amphiphile with an acid head group flanking each side of the hydrocarbon chain (hexadecanedioic acid, a bola-amphiphile) can form vesicles according to various preparation methods [66]. This last amphiphile type could be used to stably anchor the protocellular information component. One could enrich the environment with some of these single-chain amphiphiles and explore which evolutionary innovations the protocell can discover.

One could also explore a protocell variant where the ruthenium bipyridine is neither covalently attached to the DNA nor anchored to the container interface, but provided in solution. However, the protocell will employ an asymmetric DNA component where the dangling end can function as an aptamer that binds the ruthenium complex non-covalently. A machine learning method [67] could be used as a SELEX procedure that enhances the aptamer's binding affinity. In this set-up, it should be possible to feed the system with ruthenium bipyridine derivatives (or even entirely different photo-sensitizers) that differ in their binding affinities such that the protocell is forced to evolve new properties (binding of new components) rather than merely optimizing its metabolic rate.

5. Conclusion

We present the state of the art for the assembly of a particular artificial, physico-chemical minimal living system, and we provide a theoretical investigation of the main missing sub-process: a non-enzymatic self-replicating nucleotide (DNA) system. We discuss the analytically predicted template (information system) replication rate and show its critical dependence on strand length, temperature, ligation rate and detailed nucleoide hybridization energies.

We further present a simple design principle, based on dynamical hierarchies, that has been verified in both our experimental and computational systems. Observable novelty is generated as a result of self-assembling and self-organizing processes in a fixed environment with a constant throughput of free energy. Additional novelty may be observed as the complexity of the environment is increased. We, therefore, conjecture that (carefully) increasing the richness of the environmental system is a viable method for enhancing the evolutionary potential for any system in addition to possible variations (e.g. mutations) of information that encode for functionalities as, for example, in modern biological systems.

Competing interests

We declare we have no competing interests.

Funding

The authors acknowledge partial financial support from the EC sponsored projects MICREAgents and Synenergene as well as the University of Southern Denmark. S.R. is in addition grateful for partial financial support from the Santa Fe Institute during visits there.